KOSEN 한인과학기술자네트워크

Research

Our research interests lie in the areas of image processing and computer vision.

Increasing demands for multimedia data has brought significant interest in the digital transmission and storage of image and video signals. But, image and video signals often require huge storage space and high transmission bandwidth, which necessitates the development of efficient compression techniques. In 1980 and 1990s, we performed extensive studies on vector quantization (VQ) and discrete cosine transform (DCT), which are effective tools to describe still images in a compact way. Recently, we have focused in the development and application of hybrid technique of DCT and motion-compensated prediction, which is basis for most video coding standards, such as MPEG-1,2,4 and H.261/3/L. Our research interests also include wavelet-based compression, robust transmission of image and video signals over networks, and digital watermarking for copyright protection.

We are also researching theoretical and practical aspects of several computer vision problems. We have researched various techniques for 2D/3D object recognition, such as the relaxation, eigenvector matching, appearance-based recognition, and geometric invariant methods, and implemented the real-time processing systems for the developed algorithms. Another main topic is to acquire the shapes of objects from multiple input images or range data obtained by laser scanners. To this end, we have developed stereo matching techniques and registration/integration/modeling systems. In addition to these topics, we are also working on face detection and recognition, image retrieval based on color indexing, and voxel-based shape description.

In 2000, SPL was designated as a National Research Laboratory for 3D visual information processing technology. We are at the stage of combining our experiences on image processing and computer vision technologies to develop a state-of-art processing system for 3D visual data. Specifically, the goal is to develop a unified real-time system for 3D visual data acquisition, compression, and transmission.

-Medical Imaging (sponsored by KOSEF)

-Automatic organization of object-wise visual information from image collections (sponsored by MKE)